AI training for frontier models remains concentrated in hyperscale data centers, while decentralized GPU networks are increasingly targeting inference and routine workloads as a lower-cost option. Executives from Theta Network, Ovia Systems, Fluence and Salad Technologies describe how consumer and gaming GPUs can handle open-source models, data processing and geographically distributed tasks that do not require tightly synchronized hardware. These networks are positioned as a complementary layer in the AI computing stack rather than a replacement for centralized training clusters.

OpenAI is reportedly working on a new social platform built around proof-of-personhood to tackle bot activity, exploring biometric verification tools such as Face ID and World's iris-scanning technology. As interest in biometric digital identity rises, the Worldcoin (WLD) token, linked to Sam Altman's World project, surged more than 16% to $0.53 within hours.

In late January 2026, Microsoft introduced Rho-alpha, a robotic foundation model that coordinates two arms using vision, language, and tactile sensing. The system adjusts its grip based on physical feedback for tasks like inserting plugs and assembling parts, and will first roll out via an Early Access Program before becoming available on the Microsoft Foundry platform. Manufacturers and logistics firms can then adapt Rho-alpha with their own data for repetitive handling and small-batch assembly work.

OpenAI is reportedly in the early stages of developing a biometric-based social network that aims to block automated bots by enforcing proof-of-personhood logins. The project plans to use World's iris-scanning Orb devices and tools like Apple Face ID for human verification, while regulatory bans in Hong Kong and Portugal highlight privacy concerns over biometric data. World (WLD) surged about 27% on Thursday as investors reacted to reports linking the token's ecosystem to this experimental platform.

SpaceX is in early talks to merge with xAI and has also weighed a broader combination that could involve Tesla, according to Reuters. The discussions tie into a potential public listing expected later this year, while prediction market odds imply roughly a 50% chance of a SpaceX–xAI deal by mid‑2026 versus about 16% for a Tesla–xAI combination.

Between October 2025 and January 2026, security researchers reported that around 175,000 private AI servers running Ollama were exposed to the public internet, with about 23,000 of them remaining persistently online. These unprotected systems span 130 countries and are being hijacked by criminals to steal computing resources, bypass security safeguards, and potentially access internal data using techniques like SSRF, scanning, and prompt injection. Around 48% of the exposed hosts allow tool-calling, and most use a common model file format that could let attackers crash thousands of servers with a single exploit.

Nvidia's H200 chip sales in China remain paused as of January 29, 2026, with Beijing yet to approve imports and the U.S. export license still under review. The company reports no orders from Chinese customers so far, though corporate interest is strong. CEO Jensen Huang said Nvidia will act quickly once approvals are granted.

On January 29, 2026, Tesla said it will invest $2 billion in xAI and reiterated that Cybercab production is slated for 2025. The company reported a 3% full-year revenue decline to $94.83 billion for 2025, outlined more than $20 billion in 2026 capital spending, and will discontinue Model S and Model X.